AI use cases: Preparing for your first AI pilot project

In our previous article we went over six things you should do before you kick off your first artificial intelligence (AI) project, including everything from evaluating your organization's AI readiness to developing a culture of innovation. Here we'll go over some ways to prepare for and launch your first generative AI (gen AI) pilot project.

Download the "Open the future" e-book

When you think you're ready to embark on your first AI app development project, you'll first have to identify, select and prioritize some AI use cases. To do this effectively, you should:

• Start with a problem, not an AI solution
• Brainstorm problems that gen AI might help solve
• Scale up what you're already good at
• Apply AI to small, specific tasks
• Make sure you have the data you need

Let's take a closer look at how these can help your AI projects succeed.

Start with a problem, not an AI solution

As interest in AI has exploded and everyone is looking for ways to leverage these new technologies in their organizations, a lot of AI projects fail because they end up being solutions in search of a problem.

Let's say someone wants to play with some AI tools, so they kick up an experiment that seems like it could be useful. A few other people get excited about the project, and end up spending a bunch of time, money and resources developing a fully-fledged AI application. Reality sets in at launch, however, because the project didn't begin with a well-defined and clearly-scoped problem. The AI application doesn't serve any particular purpose and is eventually abandoned.

This can be an easy trap to fall into, but it can often be avoided by clearly identifying real business problems or user pain points that gen AI can help address.

Brainstorm problems that gen AI might help solve

Now that you've built a cross-functional AI team, have them get together to brainstorm some ways that AI could make a real impact in your organization.

Here are some questions they could ask to get started:

• How could gen AI help us enhance productivity, reduce costs or improve customer experience?
• What repetitive or time-intensive tasks or processes could we use gen AI to automate?
• How could gen AI personalize experiences for our customers?
• Are there untapped opportunities that gen AI could help us unlock?
• Do we have data that we've been collecting for years that gen AI could help us take better advantage of?
• Are there existing risks or unknowns that gen AI could help us better understand or mitigate?
• What business value or organizational strengths could gen AI help us augment or expand?

Scale up what you're already good at

Of course, when you think about AI models, you have to think about what you actually want to model.

For example, if your customer service is excellent and your customers rave about it, try to figure out how you can model it so AI can help you scale up that quality of service. Can you use AI to automate things so every customer gets the absolute best of your customer service, every time, regardless of how or when they contact your organization? Are there knowledge bases inside your company that can be modeled to help your customer service agents themselves respond to inquiries?

In what other ways does your organization excel that AI could help automate and scale up?

Apply AI to small, specific tasks

Many AI projects fail because the AI technology is being asked to do too much all at once. Gen AI applications are most effective when given clearly defined objectives and boundaries.

Simpler, specific tasks also reduce the chance of the AI producing errors or hallucinations. AI is unable to understand context (without specific prompting or augmenting the model's trained data) and is poor at making broad generalizations, which can lead to difficulties when attempting to manage more complex output or tasks.

For example, instead of trying to use AI to generate new long-form original content, you might be better off developing a purpose-built AI model trained on your corporate style guide that helps your human writers more efficiently edit their work.

Make sure you have the data you need

An AI model is only as good as the data used to train it. Coming up with the next killer AI app idea is all well and good, but it will amount to nothing if the data you need either doesn't exist, or is low quality.

If you have a specific AI use case in mind, work with your cross-functional AI team to brainstorm what data would be needed to effectively train a model for that use case. Then assess your available data sources to see if that data exists, is easily accessible and is of high enough quality to be of use.

Guidelines for launching an AI pilot project

Our "Open the future: An executive's guide to navigating the era of constant innovation" e-book goes into more detail, but when you're ready to launch your first AI pilot project here are some things to keep in mind.

Choose your AI platform wisely

With the speed of change and development in the AI industry, you can help future-proof your AI projects by selecting AI solutions built on open platforms that provide flexibility and choice.

Start with a small AI project

We've already discussed the importance of having a clearly defined and well-scoped AI use case, but it bears repeating—until your team or organization has more experience with developing AI models and gen AI applications, it's wise to start with small and focused.

Measure, iterate and monitor

Be sure to define your success conditions up front so you know what you're actually trying to do and whether you're succeeding. Make sure you have metrics in place so you can measure your project's impact, and check in on progress frequently. As you learn what is and isn't working, this will help you make informed decisions about how your AI application can be improved as development continues.

Also, keep an eye out for "data drift" and make sure that the data your model was trained on is consistent with the data seen in production. This sort of inconsistent data can cause your model's behavior to change over time.

Partner wisely

As we at Red Hat like to say, "No one innovates alone." Take the time to identify and select trusted partners that will help your chances of success and that you can grow with into the future.

How Red Hat can help

Red Hat's long history of open source leadership and upstream contributions is continuing apace in this new era of AI. Not only are we on the leading edge of open AI platforms, we're continuing to provide vendor-agnostic solutions that help customers take full advantage of the hybrid cloud for deploying, hosting and scaling their AI applications.

You can read more about the Red Hat advantage in this e-book, but here's a rundown of what we offer:

• A history of open source leadership
• The freedom and flexibility to innovate without constraints
• AI that is accessible and adaptable for all
• Deep roots in the wider community and ecosystem
• Industry leading expertise and experience
• Enterprise-grade support and security

Red Hat AI

When combined, Red Hat Enterprise Linux (RHEL) AI and Red Hat OpenShift AI provide a fully-featured AI model development, tuning and hosting platform that can be deployed, managed and scaled the same way you would deploy, manage and scale any containerized application.

RHEL AI includes:

• Granite open source-licensed LLMs that are distributed under the Apache 2.0 license with industry-leading transparency into training datasets and model weights
• A supported, lifecycled distribution of InstructLab that provides a scalable, cost-effective solution for enhancing and fine-tuning LLM capabilities
• A bootable image of RHEL which includes popular AI libraries such as PyTorch, and an array of hardware-optimized accelerators
• Enterprise-grade technical support provided by Red Hat
• Open Source Assurance legal protections available to all Red Hat customers with active, paid Red Hat software subscriptions

You can learn more about RHEL AI here: RHEL vs. RHEL AI: What's the difference?

Talk to a Red Hatter

With the technological foundation of Linux, containers and automation, Red Hat’s open hybrid cloud strategy gives you the flexibility to run your AI applications anywhere you need them.

If you would like to know more, contact us and talk to a Red Hatter directly.

Learn more about Red Hat AI

• Red Hat Enterprise Linux AI
• Red Hat OpenShift AI
• AI: Red Hat's vision for an open source future
• E-book: Top considerations for building a foundation for generative AI
• E-book: Building a strong operational foundation for Generative AI and beyond